rethinking european elections: the importance of regional … · 2018. 5. 4. · 6 denmark...

Rethinking European Elections: The Importance of RegionalSpillover Into the European Electoral Arena*

ARJAN H. SCHAKELMaastricht Universiy

AbstractThe European Union has developed into a multilevel electoral system in which elections take placeat regional, national and European tiers of governance. Election scholars have taken up an interestin European and regional elections and the dominant approach to study these elections is thesecond-order election model. This model has generated important insights but it also faces limita-tions. A multilevel party system perspective exposes several vertical and horizontal interactionsbetween electoral arenas. An empirical analysis on second-order effects in European elections heldin 217 regions and 11 countries reveals that regional authority, regional identity and electoraltiming of regional vis-à-vis European elections significantly alters the extent and way in whichnational politics flows into the European electoral arena. These results provide strong evidence thatregionalization has significantly transformed European elections but this impact only comes to thefore once one adopts a multilevel party system perspective.

Keywords: European elections; regional spillover; second-order election model; multilevel electoralsystem

Introduction

Representative institutions in Europe have dramatically changed over the past decades.The number of Member States holding European elections increased from 9 in 1979,when the first direct election to the European Parliament was held, to 28 for the latestEuropean election in 2014. The increasing scope of European elections was matchedby an increase in powers for the institutions of the European Union (EU). Decision-making powers have been shifted upwards from national government to the EU whichnowadays decides on important policies such as agriculture, external trade, fiscal andmonetary union, and the internal market. The story of a widening and deepeningEurope is probably familiar to most scholars but less well known is that a similar risein electoral democracy has taken place at the regional level.1 Today 19 out of 28 EUMember States hold elections for regional assemblies and, since 1970, no less than15 out of 28 EU-Member States have introduced regional elections (Dandoy andSchakel, 2013; Schakel, 2017). The establishment of regional electoral institutions alsoincluded a significant reallocation of authority downwards from national to regional

*I would like to thank JCMS’ anonymous referees for excellent comments and feedback, as well as the symposium editorsHeather A.D. Mbaye and Michaël Tatham for advice on a draft version. I would also like to thank the Hanse-Wissenschaftskolleg in Delmenhorst, Germany for providing financial support as well for being gracious hosts duringmy research leave which enabled me to collect European election data.1 Regional government is defined as a tier between local and national government with an average population of atleast 150,000 citizens (Hooghe et al., 2016). This population threshold is similar to the lower boundary of the NUTS-3classification (Nomenclature des unités territoriales statistiques) which are defined as regions with an average populationsize between 150,000 and 800,000 citizens.

JCMS 2018 Volume 56. Number 3. pp. 687–705 DOI: 10.1111/jcms.12712

© 2018 University Association for Contemporary European Studies and John Wiley & Sons Ltd

government. Many regions can tax corporate and personal income, can borrow freely onthe financial market, and have competencies in key policies such as immigration,education, culture and welfare (Hooghe et al., 2016).

The upshot is that the EU has transformed into a multilevel political system wherebyvoters have multiple opportunities to voice their preferences about policy and to electtheir representatives.2 The proliferation of supra- and sub-national elections has notescaped the attention of election scholars and they developed the second-order election(SOE) model. Although this model has generated useful insights into European andregional electoral dynamics, it also faces important limitations. The SOE-modelmatches regional or European election results to the outcomes of previously heldnational elections. This comparison overlooks the possibility that regional politics mayalso spillover into the European electoral arena. In this symposium article, I argue thata multilevel party system conceptualization is required for a comprehensive understand-ing of how electoral dynamics in one electoral arena may spillover into another electoralarena. An empirical analysis of European election results in 217 regions in 11 EUMember States shows that SOE-effects are affected by regional authority, regional iden-tity and the timing of regional elections vis-à-vis European elections. The results revealthat regional politics significantly modifies the extent and way in which national politicsflows into the European electoral arena. The results also expose a need for new conceptualthinking and in the discussion I lay out the most important linkages to consider when oneanalyses electoral outcomes in multilevel electoral systems.

In the next section I briefly discuss the downward decentralization trend from national toregional government and I show that this included a significant proliferation of regional elec-toral institutions. In the section II, I point out the limitations of the SOE-model regarding elec-toral dynamics in multi-tiered systems. In section III, I develop a conceptualization of theEuropean multilevel party system and I introduce four hypotheses on how regional politicsmay impact European election outcomes. Section IV introduces data andmethod and presentsthe empirical analysis. The final section discusses the implications of the presented results.

I. The Proliferation of Regional Electoral Institutions

A detailed measurement for the authority exercised by regional government is providedby the regional authority index (Hooghe et al., 2016). This fine-grained measurementtraces decentralization in the 28 EU Member States on annual basis from 1950 to 2010along two dimensions. Self-rule is the authority that a subnational government exercisesin its own territory. Shared rule is the authority that a subnational government exercises inthe country as a whole.3 Scores range from 0 (no regional government) to a maximum

2 See Tatham and Mbaye’s contribution to this symposium.3 Both self-rule and shared rule consist of five sub-dimensions. Self-rule measures the extent of independence of a regionalgovernment from national government (institutional depth), the range of a regional government’s authority over policy in itsjurisdiction (policy scope), the authority a regional government has over taxation within its own jurisdiction (fiscal auton-omy), the authority of a regional government to borrow on financial markets (borrowing autonomy), and whether a regionalgovernment is endowed with representative institutions (representation). Shared rule measures the extent to which aregional government co-determines national policy-making (law-making), the extent to which a regional government co-determines national executive policy in intergovernmental fora (executive control), the extent to which a regional govern-ment co-determines how national tax revenues are distributed (fiscal control), the extent to which a regional governmentco-determines the restrictions placed on borrowing (borrowing control), and the extent to which regional governmentcan initiate or constrain constitutional reform (constitutional reform).

Arjan H. Schakel688


score of 30 (full authority). Figure 1 displays average scores for old and new western EUMember States and for post-communist EU Member States.4 Although starting in differ-ent time periods and from different initial levels, the trend in regional authority is similaracross Europe: authority for regional government is increasing.

The incidence of reform has been especially notable for representative institutions. In1950, only 9 out 28 EU Member States held regional elections, 5 across the statewideterritory5 and 4 in specific territories.6 Between 1970, when Italy introduced electionsfor the regioni a statuto ordinare, and 2001, when Slovakia held the first elections tothe assemblies for samosprávné kraje, 22 reforms have been implemented. Hence, duringthis period, regional representative institutions have been, on average, introduced everyone and a half years. None of these reforms have been reversed nor have the original ninecountries disestablished electoral institutions for their regions.7 Most European citizenscan elect regional and European representatives alongside local and national deputies.In 2017, regional parliaments rule over 420 million people and the European Parliamentrepresents 506 million people.

II. Methodological Nationalism Inherent to the Second-Order Election Model

The proliferation of European and regional electoral institutions has opened up two newfields for comparative inquiry: European (Reif and Schmitt, 1980) and regional elections(Jeffery and Wincott, 2010). The pre-dominant approach to study regional and Europeanelections is the second-order election (SOE) model (Hix and Marsh, 2007, 2011; Pallarésand Keating, 2003; Tronconi and Roux, 2009). In a SOE turnout is low and parties in

Figure 1: Development in Regional Authority since 1950

Source: Hooghe et al. (2016).

4 Old western EU member states: Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Luxemburg,Netherlands, Sweden, and the United Kingdom. New western EU member states: Cyprus, Greece, Malta, Portugal, andSpain. Post-communist EU member states: Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania,Poland, Romania, Slovakia, and Slovenia.5 Austria (Länder), Belgium (provincies), (West-)Germany (Länder), Netherlands (provincies), Sweden (län).6 Denmark (Føroyar), Finland (Åland), Italy (regioni a statuto speciale), United Kingdom (Northern Ireland).7 Elections in Northern Ireland were suspended between 1977–99 and 2002–07 and a pilot program in Finland whichintroduced elections to the provincial assembly of Kainuu in 2005 was suspended in 2012.

Rethinking European elections 689


national government lose vote share, whereas opposition, small and new parties gainvotes (Reif and Schmitt, 1980). The assumption is that this electoral behaviour is broughtabout by the fact that there is ‘less at stake’ in SOEs compared to first-order, nationalelections (Marsh and Mikhaylov, 2010). Second-order parliaments do not havecomparable decision-making and executive oversight powers as national parliaments(Reif, 1985). As a result, voters do not care to turn out (Fauvelle-Aymar and Stegmaier,2008; Franklin and Hobolt, 2011) except for discontent voters who voice their dissatisfac-tion by punishing parties in government while favouring opposition and small parties(Oppenhuis et al., 1996; Van der Eijk et al., 1996). Another outcome is that mediacoverage tends to be low and political parties tend to invest fewer campaign resources(De Vreese et al., 2006; Weber, 2007).

The SOE-model has generated useful insights but by taking this model as a startingpoint scholars also incorporate the assumption that regional (and European) electionsare subordinate to national politics.8 The nation-state constitutes the main unit ofanalysis whereas the region is neglected as a unit for political analysis. This is criticizedby Jeffery and Wincott (2010) as ‘methodological nationalism’: election scholars too eas-ily assume that the nation-state and, by extension, the national election, is the only or themost dominant arena for political contestation. An example will help to clarify how thisassumption may lead to shortcomings in our understanding of electoral outcomes.

The SOE model compares vote shares between European and preceding nationalelections. Parties in national government should lose vote share whereas oppositionparties should win vote share. The Scottish National Party (SNP) is classified by SOEscholars as an opposition party because the party is not part of executive government atthe UK-level and a vote share win for the SNP in European elections is interpreted as aSOE-effect. However, scholars working on Scottish nationalism and/or Scottish electionswill probably very quickly point out that the vote share gain for the SNP may result fromvoters who positively evaluate the SNP’s performance in government at Holyrood and notbecause the party was in opposition in Westminster (Schakel and Dandoy, 2013). This‘territorial effect’ is not on the ‘radar’ of election scholars when the SOE-model is takenas a theoretical starting point (cf. Keating, 1998, p. ix, 2009).

There is an increasing body of evidence that the region has become much moreimportant as a locus for social and political life over the past four decades. There iscompelling evidence that shows that inter-regional policy variation is widening (forexample, McEwen and Moreno, 2005) and that the number of regional (or non-statewide)political parties contesting regional elections is growing (Jeffery, 2010; Massetti andSchakel, 2017). So far, this evidence has led election scholars to explore when and whereregional elections are second-order and they discovered significant variation in SOE-effects across regional elections (Dandoy and Schakel, 2013; Hough and Jeffery, 2006;Pallares and Keating, 2003; Schakel, 2017). However, theoretical innovation by thisscholarship has mostly been limited to refining the SOE-model for regional elections.To fully appreciate the importance of the regional electoral arena one needs to adopt aEuropean multilevel party system perspective.

8 A more encompassing discussion and a more elaborate critique on the second-order election model can be found inSchakel and Jeffery (2013).

Arjan H. Schakel690


III. Conceptualizing the European Multilevel Party System

Swenden and Maddens (2009, p. 6) provide a useful definition for a multilevel partysystem which ‘brings together a statewide party system which emerges from statewideelections and a set of regional party systems reflecting the outcome for regional elections’.Swenden and Maddens (2009, p. 6) point out that this conceptualization invites one toconsider how regional party systems interact with each other (horizontal interaction)and with the statewide party system (vertical interaction).9 In addition, a multilevel partysystem perspective reveals that ‘methodological nationalism’ is inherent to the SOE-model. This model is concerned with a vertical ‘top-down’ interaction, that is on howdevelopments in the statewide electoral arena affect regional party systems. However,vertical interactions can also be bottom-up, that is developments in the regional electoralarena can trigger changes in the statewide electoral arena. For example, party success inregional elections may lead to electoral success at the national level because parties canuse the regional electoral arena as a ‘springboard’ to gain access to the statewide electoralarena (Brancati, 2008; Massetti and Schakel, 2013).

In addition, the SOE-model overlooks possible horizontal interactions betweenregional party systems. One example is the widening support for autonomist parties inSpain during the 1980s and early 1990s (Swenden and Maddens, 2009, pp. 8–9). Pallarésand Keating (2003) show that this was mainly a consequence of the rapid growth of theseparties in non-historic communities (such as the Canary Islands or Aragon) rather thanfurther electoral growth of autonomist parties in the historic communities (BasqueCountry, Catalonia and Galicia). In other words, the emergence of autonomist parties inthe historic communities triggered the break-through of autonomist parties in otherregional elections (a ‘snowball-effect’). Horizontal and (bottom-up) vertical interactionsmay also unfold simultaneously. The Greens in Austria and Germany (respectively DieGrünen–Die Grüne Alternative and Bündnis 90–Die Grünen) first achieved electoralsuccess and governmental representation in one or few Länder, then they gained represen-tation in other Länder, paving the way for statewide electoral success and a role in federalgovernment at a later stage.

Research has shown that European elections are second-order regarding first-order,national elections (Hix and Marsh, 2007, 2011) but the extent to which regional politicsimpacts European election results has, as far as I know, not been assessed. Distinguishingtheoretically and empirically between regional and national spillover into the Europeanparty system falls beyond the scope of this symposium article.10 Instead, I will focuson the question of whether regionalization has an indirect transformative effect, thatis, whether regional politics modifies the extent to which national politics spills overinto the European electoral arena. When this modifying effect is shown to exist

9 Swenden and Maddens (2009, p. 6) also point out that integration of the party system is also a focal point for electionscholars: ‘The integration of a multilevel party system depends on the extent to which the same parties are represented atthe various levels of the system and how evenly balanced their support is across the levels and regions of the state’. In thissection, I focus on horizontal and vertical interactions.10 Studying the direct impact of regional politics on European election results requires a comparison between European andpreviously held national and regional elections whereby vote share swings are calculated for government, opposition, noseat, and new parties in relation to both the national and regional electoral arenas. Furthermore, such an analysis also re-quires an empirical strategy which allows for parsing out national from regional spillover. When it is shown that regionalpolitics impacts indirectly on European election results it is worthwhile to embark on this data-intensive endeavour.



empirically it suggests, at the very least, that the SOE-model for European electionsneeds to be refined to accommodate territorial effects. I hypothesize four possible waysin which regional politics may impact the magnitude of SOE-effects in Europeanelections.

First, SOE-effects in European elections decrease as regional authority increases.Voters from within powerful regions are accustomed to making decisions in two equallyimportant elections. These voters have developed ways to access information aboutcandidates, parties and their policies for both regional and national elections and for themit is relatively easy to transfer similar behaviour into a third (European) electoralarena.11 This is not the case for a voter from within less powerful regions who has ‘tolearn from scratch’. These voters may rely on ‘short-cuts’ such as cues arising from the na-tional electoral arena. When another non-important election is held, these voters simplytransfer their SOE-behaviour from one electoral contest into another and they do not inves-tigate how far candidates and parties and their issues and policies differ across SOE-arenas.

Second, SOE-effects in European elections are smaller in regions with strong regionalidentities. Studies have shown that regional parties are consistent and convinced sup-porters of European integration (Jolly, 2007), in particular those which are secessionistand aim for an independence for their region.12 Although there is evidence for a‘eurosceptic turn’ among these parties in the 2000s (Elias, 2008, Massetti, 2009), overallthe regional party family is still europhile (Bakker et al., 2012; Szöcsik, 2013). Inaddition, EU regional policy has reinforced the region as a relevant decision-makinglevel which induces regional parties to constitute themselves as competent actors torepresent the region in Brussels to attract EU subsidies to the region and to make sure thatregional interests are voiced at the EU level (De Winter et al., 2006, p. 202).13 Regionalparties mobilize their voters to turn out and vote for them in European elections and, as aresult, voters are less inclined to base their vote choice on cues arising from the nationalelectoral arena.

Third, SOE-effects increase when a regional election precedes a European electionbut, fourth, SOE-effects decrease when regional and European elections are heldconcurrently. An electoral timing effect can be expected while it relates to the basicassumption of the SOE-model, i.e. what is at stake in elections determines the extentto which we may observe SOE-effects. When elections are held on the same date,the stakes for political parties and candidates increase resulting in more intensivecampaigns and increased media attention which makes it easier for the voter to accessinformation (Cox and Munger, 1989). Furthermore, simultaneous elections lower thecosts for voters to cast a vote because the fixed costs of turning out can be divided overmultiple elections (Geys, 2006). When regional elections are more important thanEuropean elections one may also expect that voters base their European vote choiceon cues arising from the regional electoral arena instead of or in addition to the nationalelectoral arena. In non-simultaneous elections, the involved stakes for parties andcandidates are relatively lower and campaign efforts and media attention diminish as

11 See also Léon’s contribution to this symposium.12 See also Cetrà and Liñeira’s contribution to this symposium.13 See also Tatham’s contribution to this symposium.

Arjan H. Schakel692


a result (Henderson and McEwen, 2010). Additionally, the relative costs of voting arehigher since elections are multiplied. SOE-effects will increase to the extent that thestakes decrease and this may especially happen when one type of SOE is precededby another type of SOE, such as when a regional election has been held before aEuropean election.

IV. Data, Method and Empirical Analysis: Regional Spillover into the EuropeanElectoral Arena

The empirical analysis will focus on 11 EU Member States which hold regional, nationaland European elections (Table 1).14 The development of a multilevel electoral systemdiffers across this set of countries. Belgium, Denmark, France, Germany, Italy, theNetherlands and the United Kingdom have held European elections ever since 1979whereas Greece, Spain, Sweden and Austria followed in 1981, 1987, 1995 and 1996,respectively.15 A country’s experience in holding regional elections also varies consider-ably. Austria, West Germany, special statute regions in Italy, and the Netherlands haveheld regional elections since 1950. Italy (ordinary regions) and Sweden introduced

14 The dataset is part of an on-going large data collection project on regional, national and European election results at theregional level for all 28 EU Member States for elections held since 1979 or the first time a European election was held in acountry.15 The analysis excludes these four European elections because they were held in one country only.

Table 1: Included Countries and Regions, Number of Parties, Regions, and Included European andNational Elections

Country Parties Regions European National

First Last First Last

Austria 21 Länder 9 1999 2014 1995 2013Belgium 47 Communities and Regions 4 1979 2014 1978 2014Denmark 19 Amter 15 1979 2004 1977 2001

Regions 5 2009 2014 2007 2011France 43 Régions 22 1979 2014 1978 2012Germany 28 Länder (West) 10 1979 2014 1978 2013

Länder (East) 6 1994 2014 1990 2013Greece 41 Nomoi 48 1984 2004 1976 2004

Peripheria 13 2009 2014 2007 2012Italy 109 Regioni ordinare 15 1979 2014 1979 2013

Regioni speciale 4 1979 2014 1979 2013Provincia speciale 2 1979 2014 1979 2013

Netherlands 38 Provincies 12 1979 2014 1977 2012Spain 82 Comunidades (non-historic) 14 1989 2014 1986 2011

Comunidades (historic) 3 1989 2014 1986 2011Ciudiades 2 1989 2014 1986 2011

Sweden 22 Landstinge 21 1999 2014 1998 2010UK 60 Devolved entities 4 1979 2014 1979 2010

Regions 8 1979 2014 1979 2010



regional elections in the 1970s; France, Greece and Spain established regional electoralarenas in the 1980s; Germany re-instated elections for the Eastern Länder in 1990; andBelgium and the United Kingdom introduced regional elections later in the 1990s.

The 11 EU Member States also differ regarding the three key independent variables ofinterest: regional authority, regional identity and electoral timing. Belgium, Austria andGermany are federal states; Italy, Spain and the United Kingdom are regionalized states;and Denmark, France, Greece, the Netherlands and Sweden are (decentralized) unitarystates (Loughlin et al., 2011). In addition, regional authority is symmetrical in Austria,Denmark, France (except for Corsica), Germany, Greece, the Netherlands and Swedenbut is asymmetrical in Belgium, Italy, Spain and the United Kingdom. Regional authorityalso varies across time. All eleven EU Member States implemented multiple regional re-forms and Denmark and Greece replaced one tier of regional governance for another. Inall eleven EU Member States one can find regions where regional parties compete in elec-tions but they tend to be nearly absent in Denmark and Greece (Massetti and Schakel,2017). Finally, national and regional elections tend to be held non-simultaneously withEuropean elections, except for some instances.16

Second-order election effects are assessed by observing vote share swings betweenEuropean and previously held national elections.17 A party enters the dataset when itreceives a regional vote share of at least 1 per cent in a European or national election.Regional vote shares for European elections are subtracted from regional vote shareswon in preceding national elections. Vote share swings are summed for five types ofparties (Schakel, 2015):

• Government parties held executive office at the national level at the time of theEuropean election;

• Main and other opposition parties won seats in national parliament but are not part ofthe executive government. I differentiate between the main opposition party that wonmost parliamentary seats and other (smaller) opposition parties because the mainopposition party can be expected to attract the protest vote18;

• No seat parties contested a national election but did not manage to win a parliamentaryseat;

• New parties did not participate in the previously held national election but won voteshare in the European election.

The regional authority index provides scores for self-rule and shared rule (Hoogheet al., 2016) which are entered separately into the models to trace the differences between,on the one hand, regions in unitary states (Denmark, France, Greece and Sweden) whichonly have self-rule, and on the other hand, regions in federal (Austria, Belgium andGermany), regionalized (Italy, Spain and the United Kingdom), and unitary decentralized

16 Simultaneous European, national, and regional elections: Belgium (1999, 2014); simultaneous European and nationalelections: Greece (1989); simultaneous European and regional elections: Belgium (2004, 2009), Spain (1999, non-historic communities), Abruzzo (2014), Andalusia (2014), Piemonte (2014), Sardinia (1994–2004), Thuringia (2004).17 Another effect is lower turnout in the second-order election compared to national elections. I do not look at turnout be-cause such an analysis requires the inclusion of several additional explanatory variables – such as financial contributions toor from the EU and public support for the EU (Matilla, 2003) – which tend not to be available at the regional level.18 Government parties are not differentiated between a main and other category because the other category remains emptyin cases of single party government which occurred frequently in France, Greece, Spain, Sweden and the United Kingdom.

Arjan H. Schakel694


states (the Netherlands) which have both self-rule and shared rule. The impact of regionalidentity is assessed by three dummy variables: a regional language dummy which scorespositive when a majority of people in a region speaks a different language from the mainnational language; a statehood variables which scores positive when a region has a historyof sovereign statehood for at least 30 years between 1200 AD and 1950; and a distancevariable which scores positive when a region is an island or archipelago and at least 30km from the mainland (Hooghe and Marks, 2016). One may expect that regional partieswill contest elections in these regions because they can mobilize the regional voter byinvoking regional identities which can be based on speaking a regional language, ahistory of independent statehood, or geographical distance from the centre.19 Twodummy variables assess the impact of electoral timing of the European vis-à-vis regionalelections. One dummy indicates when a European election is held before a regionalelection and another dummy indicates when a European election is held concurrently witha regional election.

The models include six control variables to control for national spillover in theEuropean electoral arena. SOE-effects tend to follow a parabolic path whereby vote sharelosses (gains) are largest at mid-term of the national election cycle (two years in mostcountries) and tend to be smaller the closer the European election is held to a precedingor forthcoming national election. This relationship is modelled by introducing a cycleand cycle-squared variable whereby the cycle variable is derived by taking the numberof days between a European and a previously held national election and dividing thisnumber by 365 days to obtain years. One dummy indicates simultaneous European andnational elections, another dummy indicates elections with compulsory voting, and a thirddummy signals national elections conducted under (more) majoritarian rule (all Europeanelections are held under proportional rule). Finally, the size of regions is a ratio variablewhich is derived by dividing the regional electorate by the statewide electorate (both fornational elections).20 Descriptive statistics are provided in Table A1 in the Appendix.The method for model estimation is ordinary least regression with panel correctedstandard errors (Beck and Katz, 1995, 2011) and the full results are shown in Table A2in the Appendix. The models include country and European election dummies becauseregions are clustered in countries and European elections have taken place at the samedate.21

Figure 2 displays vote share swings between European and previously held nationalelections for five party categories. Shown are averages and their 95 per cent confidenceintervals and these estimates are derived by setting the self-rule and shared rule scoresas shown in Table 2 while keeping all the other independent variables at their mean(continuous variables) or mode (dummy variables) (see Appendix Table A1) and byincluding a European election fixed effect for 2014. Countries with regions which haveonly self-rule are shown on the left-hand side whereas countries with regions which have

19 I do not classify regional parties as a separate category because this would interfere with the categorization of parties. Aregional party can be in government or in opposition and can be a new, or no seat. When party categories overlap one cannotattribute a vote share loss or gain to one single party category.20 The number of valid votes is taken in case the size of the electorate is missing.21 The results appear to be robust across various model specifications, i.e. models which include an autocorrelation coeffi-cient, fixed effects (by region) models, and mixed effects linear regression models whereby aggregated vote share swingsare clustered in regions which, in turn, are clustered by country. These results are available upon request.



Figure 2: The Combined Effect of Self-rule and Shared Rule on the Vote Share Swing betweenEuropean and Previously Held National Elections for Five Categories of Party.

Note: Estimates are based on the results shown in Table A2 in the Appendix.

Arjan H. Schakel696


both self-rule and shared rule are shown on the right-hand side. The countries are orderedon the x-axis according to their self-rule and shared rule scores from low (left) to high(right) (Table 2).

Figure 2 reveals that European elections are second-order across all countries butSOE-effects vary according to the form and extent of regional authority. Governmentparties tend to lose more vote share in regions with only self-rule and they lose lessvote share in regions which have both self- and shared rule. In contrast to SOE-model expectations, the main opposition party loses vote share and they especiallydo so in regions with shared rule. Smaller opposition parties win vote share and voteshare gains tend to increase according to the extent regions obtain self-rule butdecrease to the extent regions acquire shared rule. No seat parties win vote share,especially in regions with less authority. New parties also win vote share, whichgiven their operationalization is not surprising, and gains increase to the extentregions acquire self-rule but tend to decrease to the extent regions obtain shared rule.Taken together, these results strongly indicate that the form and extent of regionalauthority modifies the extent to which national politics spills over into the Europeanelectoral arena.

Figures 3A, 3B and 3C display the impact of regional identity and electoral timing.Shown are the additive effects of a European election taking place in a language region(3A),22 a European election which is held before a regional election (3B), and a Europeanelection which is held at the same date as a regional election (3C). When a European

22 Among the three dummies (language, statehood and distance), only the language dummy attains statistical significance.This result corroborates that language is an important marker for regional identity (Laitin, 2000, p. 144) that may subse-quently impact voting behaviour.

Table 2: Regional Authority Index (RAI) Scores in 11 EU Member States.

Country Self-rule Shared rule RAI-score

UK Regions-a UK regions before 1994 and in 2014 0 0 0GRE<19991 Greece before 1999 1 0 1UK Regions-b UK regions between 1994 and 2014 5 0 5DEN>2009 Denmark since 2009 7 0 7FRA>19892 France since 1989 10 0 10SWE<20043 Sweden before 2004 13 0 13UK-WAL Wales in the UK 9 6.5 15.5NET4 Netherlands 10 7.5 17.5ITA>19995 Italy since 1999 15 3 18UK-SCO Scotland in the UK 14 6.5 20.5AUT6 Austria 14 9 23GER Germany 15 11 26

Notes:1 Greece before 1999 also represents France in 1979 (2+0).2 France since 1989 also represents Greece since 1999 (9+0) and Italy before 1999 (10/11+0/1).3 Sweden also represents Sweden since 2004 (12+0) and Denmark before 2009 (12+0).4 Netherlands also represents Wallonia and Flanders in Belgium between 1989 and 1994 (11+9).5 Italy since 1999 also represents Brussels in Belgium since 2004 (13+5).6 Austria also represents Spain (14+9.5) and Flanders and Wallonia in Belgium since 2004 (14+10).



election takes place in a region where a majority of people speaks a different languagefrom the main national language, new parties benefit from the loss of government parties.Contrary to expectations, government parties lose more vote share, and smalleropposition and no seat parties win more vote share, when a European election precedesa regional election. The hypothesis concerning concurrent regional and Europeanelections is confirmed while in these instances one may find opposite SOE-effects, i.e.government and no seat parties win vote share but the main and other opposition partieslose vote share. The results for regional identity and electoral timing also strongly suggestthat regional politics modifies the extent to which national politics flows into theEuropean electoral arena.

Discussion: Conceptual and Empirical Challenges of Studying Elections in aMultilevel Party System

The results presented in this symposium article strongly suggest that regional politicssignificantly alters the extent and way in which national politics flows into theEuropean electoral arena. It is therefore likely that regionalization has significantlytransformed the European multilevel party system. However, this insight only comesto the fore once one acknowledges that methodological nationalism inherent in theSOE-model precludes observing many interactions between electoral arenas. Theempirical analysis presented in this symposium article provides one illustration of

Figure 3: The Effects of Language Region and Electoral Timing on the Vote Share Swing BetweenEuropean and Previously Held National Elections for Five Party Categories.

Note: Estimates are based on the results shown in Appendix Table A2.

Arjan H. Schakel698


the complexities of multilevel electoral behaviour23 and, at the very least, reveals aneed to further develop the SOE-model so that its explanatory power will beincreased.

Figure 2 reveals an interesting difference between SOE-effects in European electionsheld in regions with only self-rule versus regions which also have shared rule. Governmentparties lose more, no seat parties gain less, and other opposition and new parties win morevote share when self-rule increases but losses and gains level off to the extent regionsobtain shared rule. This empirical result runs counter to the hypothesis that SOE effectsdecrease when regional authority increases. A possible explanation could be that Europeanelections obtain a ‘third-order election’ status when self-rule increases because there ismore ‘at stake’ in regional elections.24 When regions also obtain shared rule, voters andparties are accustomed to a situation in which the outcomes of a regional election impactspolitics at the national level. This induces both voters and parties to exhibit ‘first-orderelection behaviour’ in the regional electoral arena. Although European elections do notdirectly affect national policy-making to a similar extent as regional elections can, itmay be still the case that relatively more voters and parties in regions with shared ruletransfer their behaviour from the regional to the European electoral arena in comparisonto voters and parties in regions without shared rule.

Which interactions between regional, national and European electoral arenas deserveour attention? Apart from the differentiated effects of self-rule and shared rule I wouldlike to discuss three additional linkages. First, a ‘snow-ball effect’ of parties whichincrease their territorial scope across regional electoral arenas such as the Green partiesin Austria and Germany (Swenden and Maddens, pp. 8–9). A ‘snow-ball effect’ can alsooccur at a transnational scale for parties that increase their territorial scope across Europe.Possible examples are Pirate, eurosceptic, and right-wing populist parties. Second, a‘springboard effect’ (Brancati, 2008) whereby parties use the regional and/or nationalelectoral arena to gain access to respectively national and/or European elections. A‘springboard effect’ has been empirically shown for regionalist parties – defined as partieswith a self-rule demand – whose ambition is to win seats in the national electoral arenabecause that is where decisions on decentralization reforms are taken (Masseti andSchakel, 2013). Third, a ‘trickle-down effect’ whereby parties use the European and/ornational electoral arena to gain access to national and/or regional elections, respectively.Many new parties emerge in the European electoral arena (Figure 2) and we do not knowmuch about their electoral fate after their first appearance.

A multilevel party system perspective exposes a whole myriad of possible verticaland horizontal interactions between regional, national and European electoral arenas.And apart from posing a conceptual challenge, studying elections in a multi-levelelectoral system perspective also entails a huge data collection challenge (Schakeland Jeffery, 2013). It requires disaggregating electoral outcomes to the regional leveland it involves collecting information on regional electoral rules and on how regionalexecutives are formed. Instead of collecting data for 28 EU Member States, election

23 Another illustrating example of complex multilevel electoral behaviour is provided in León’s contribution to this sym-posium, where Leon shows that in federal countries partisans of the national incumbent party with negative economic as-sessments are more likely to assign responsibility to regional governments than partisans of other parties and non-partisans.But in non-federal countries there are no differences between partisans.24 I would like to thank an anonymous referee for suggesting that European elections may become third-order elections.



scholars are now confronted with a need to collect data for hundreds of regions,hundreds of parties and thousands of elections. Studying elections in multilevel partysystems is a theoretically challenging and a daunting practical enterprise but one whichis worthwhile pursuing while it has the potential to significantly advance scholarship onelections.

Correspondence:Arjan H. SchakelAssistant Professor in Research MethodsDepartment of Political ScienceFaculty of Arts and Social ScienceMaastricht UniversiyPO Box 6166200 MD MaastrichtThe Netherlands.email: [email protected]

Appendix

Table A1: Descriptive Statistics for the Dependent and Independent Variables.

Variable Mean Median SD Min Max

vote share swing government parties -6.73 -6.19 8.41 -37.88 33.31vote share swing main opposition parties -1.86 -1.69 5.92 -26.41 23.57vote share swing other opposition parties 2.09 1.55 7.30 -20.27 32.44vote share swing no seat parties 1.86 0.45 6.13 -25.01 36.55vote share swing new parties 4.55 1.78 6.60 0 37.74self-rule 9.66 10 4.63 0 15shared rule 3.08 0 4.19 0 12language region 0.12 0 0.33 0 1distance 0.05 0 0.22 0 1statehood 0.12 0 0.32 0 1EU election first 0.36 0 0.48 0 1simultaneous European and regional elections 0.03 0 0.17 0 1cycle 1.72 1.75 1.10 0 4.08cycle2 4.17 3.05 4.13 0 16.69simultaneous European and national elections 0.04 0 0.20 0 1compulsory voting 0.23 0 0.42 0 1different electoral system 0.51 1 0.50 0 1size region 0.06 0.04 0.07 0.00 0.61

Notes: The number of observations is 1,276.

Arjan H. Schakel700


Table

A2:

DeterminantsforVoteShare

Swings

BetweenEU

andPreviou

slyHeldNationalElections

forFiveTyp

esof

Party.

government

oppositio

nmain

oppositio

nother

noseat

new

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

self-rule

-0.60

0.10

***

-0.35

0.08

***

0.32

0.10

***

-0.25

0.08

***

0.73

0.08

***

shared

rule

1.14

0.23

***

0.02

0.16

-0.38

0.23

*-0.05

0.20

-0.44

0.15

***

language

-1.68

0.66

**-0.16

0.51

0.34

0.65

0.28

0.50

1.09

0.49

**distance

1.25

0.99

1.15

0.75

0.30

0.79

-0.87

0.54

-0.54

0.60

statehood

0.88

0.70

-0.51

0.52

0.17

0.60

0.25

0.39

0.26

0.41

EU

electio

nfirst

-3.15

0.55

***

-0.52

0.42

1.00

0.49

**2.79

0.35

***

0.13

0.37

sim

EU-reg

2.17

1.18

*-2.04

0.85

**-2.47

1.03

**2.80

0.84

***

-0.22

0.93

cycle

-7.15

0.77

***

5.16

0.59

***

0.09

0.76

-1.72

0.60

***

4.16

0.49

***

cycle2

1.49

0.19

***

-1.44

0.15

***

0.03

0.20

0.79

0.16

***

-1.03

0.13

***

sim

EU-nat

-5.01

1.27

***

3.51

0.90

***

0.72

1.15

-2.83

0.99

***

0.90

0.86

compulsoryvotin

g-10.64

0.85

***

0.16

0.60

7.18

0.82

***

2.70

0.65

***

0.65

0.50

diffelectoralsystem

-8.42

1.09

***

-0.79

0.77

1.45

0.97

5.60

0.97

***

1.74

0.58

***

size

region

-5.76

2.99

*1.40

2.03

-2.96

3.24

7.94

2.21

***

-2.61

2.15

EU

electio

nfixedeffects(1979=base)

1984

-0.82

0.96

3.93

0.71

***

-2.36

0.84

***

0.79

0.63

-1.52

0.61

**1989

-0.65

1.05

0.20

0.77

-2.58

0.93

***

0.69

0.67

1.29

0.68

*1994

-2.27

0.95

**1.20

0.70

*-0.92

0.84

-0.33

0.64

2.26

0.61

***

1999

-0.92

1.07

5.13

0.81

***

-0.40

0.99

0.79

0.66

-3.54

0.72

***

2004

-4.86

1.05

***

4.97

0.81

***

1.38

0.98

1.73

0.67

**-2.29

0.73

***

2009

-3.32

1.07

***

-0.59

0.80

3.80

1.01

***

3.39

0.67

***

-2.33

0.70

***

2014

-7.63

1.03

***

0.67

0.77

5.55

0.98

***

3.70

0.67

***

-1.78

0.69

**

Country

fixedeffects(A

ustria

=base)

Belgium

11.23

2.12

***

0.80

1.19

1.89

1.82

-4.06

1.43

***

-7.31

1.58

***

Denmark

6.05

2.37

**-6.54

1.55

***

-1.05

2.28

0.15

1.78

4.57

1.78

**France

7.52

2.71

***

-3.34

1.72

*6.84

2.54

***

0.34

2.14

-6.93

1.89

***

Greece

-0.02

1.64

6.24

1.05

***

4.83

1.50

***

-1.48

1.13

-10.19

1.22

***

Germany

15.87

2.61

***

-4.31

1.62

***

0.04

2.44

-3.04

2.12

-7.62

1.77

***

Italy

17.74

2.10

***

2.22

1.29

*-2.69

1.96

-2.16

1.58

-12.71

1.56

***

Netherlands

-7.19

1.53

***

-0.97

0.85

13.06

1.45

***

3.29

0.70

***

-7.67

1.14

***

Spain

-2.05

1.28

2.22

0.80

***

4.45

1.16

***

1.35

0.68

**-6.98

1.21

***



Table

A2:

(Con

tinued)

government

oppositio

nmain

oppositio

nother

noseat

new

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

bs.e.

sig.

Sweden

0.98

2.37

-3.00

1.49

**7.53

2.37

***

3.65

1.82

**-6.52

1.84

***

UnitedKingdom

-4.20

2.37

*1.01

1.51

1.30

2.39

9.01

1.92

***

-5.36

1.63

***

constant

6.31

2.41

***

-2.27

1.52

-6.18

2.32

***

-2.40

1.89

2.92

1.71

R-squared

0.46

0.41

0.36

0.51

0.56

Waldchi2

1035

***

1035

***

654

***

850

***

1213

***

Notes:*p<

0.10;**

p<

0.05;***p<

0.01.Shownaretheresults

ofPrais-W

istenmodelswith

panelcorrectedstandard

errors

(BeckandKatz,

1995,2011).The

dependent

variable

isan

aggregated

votesharesw

ingforapartycategory

for1,276electio

nsheld

with

in217regionsandeleven

EU

Mem

berStates(Table

1).

Arjan H. Schakel702


References

Bakker, R., de Vries, C., Edwards, E. et al. (2012) ‘Measuring Party Positions in Europe: TheChapel Hill Expert Survey Trend File, 1999–2010’. Party Politics, Vol. 21, No. 1, pp. 143–52.

Beck, N. and Katz, J.N. (1995) ‘What To Do (And Not To Do) with Time-Series Cross-SectionData’. American Journal of Political Science, Vol. 89, No. 3, pp. 634–47.

Beck, N. and Katz, J.N. (2011) ‘Modelling Dynamics in Time-Series-Cross-Section PoliticalEconomy Data’. Annual Review of Political Science, Vol. 14, pp. 331–52.

Brancati, D. (2008) ‘The Origins and Strengths of Regional Parties’. British Journal of PoliticalScience, Vol. 38, No. 1, pp. 135–59.

Cox, G.W. and Munger, M.C. (1989) ‘Closeness, Expenditures and Turnout in the 1982 US HouseElections’. American Political Science Review, Vol. 83, No. 1, pp. 217–30.

Dandoy, R. and Schakel, A.H. (eds) (2013) Regional and National Elections in Western Europe.Territoriality of the Vote in the Thirteen Countries (Houndmills: Palgrave Macmillan).

De Vreese, C.H., Banducci, S.A., Semetko, H.A. and Boomgaarden, H.G. (2006) ‘The NewsCoverage of the 2004 Parliamentary Election Campaign in 25 Countries’. European UnionPolitics, Vol. 7, No. 4, pp. 477–504.

De Winter, L., Gomez-Reino, M. and Lynch, P. (eds) (2006) Autonomist Parties in Europe:Identity Politics and the Revival of the Territorial Cleavage (Barcelona: ICPS).

Elias, A. (2008) ‘From Euro-Enthusiasm to Euro-Scepticism? A Re-Evaluation of MinorityNationalist Party Attitudes Towards European Integration’. Regional and Federal Studies,Vol. 18, No. 5, pp. 557–81.

Fauvelle-Aymar, C. and Stegmaier, M. (2008) ‘Economic and Political Effects on EuropeanParliamentary Electoral Turnout in Post-Communist Europe’. Electoral Studies, Vol. 27,No. 4, pp. 661–72.

Franklin, M.N. and Hobolt, S.B. (2011) ‘The Legacy of Lethargy: How Elections to the EuropeanParliament Depress Turnout’. Electoral Studies, Vol. 30, No. 1, pp. 67–76.

Geys, B. (2006) ‘Explaining Voter Turnout. A Review of Aggregate-Level Research’. ElectoralStudies, Vol. 25, No. 4, pp. 637–63.

Henderson, A. and McEwen, N. (2010) ‘A Comparative Analysis of Voter Turnout in RegionalElections’. Electoral Studies, Vol. 29, No. 3, pp. 405–16.

Hix, S. and Marsh, M. (2007) ‘Punishment or Protest? Understanding European ParliamentElections’. The Journal of Politics, Vol. 69, No. 2, pp. 495–510.

Hix, S. and Marsh, M. (2011) ‘Second-Order Effects Plus Pan-European Political Swings: AnAnalysis of European Parliament Elections Across Time’. Electoral Studies, Vol. 30, No. 1,pp. 4–15.

Hooghe, L. and Marks, G. (with Schakel, A.H., Chapman-Osterkatz, S., Niedzwiecki, S. andShair-Rosenfield, S.) (2016) Community, Scale, and Regional Governance: A PostfunctionalTheory of Governance, Volume II (Oxford: Oxford University Press).

Hooghe, L., Marks, G., Schakel, A.H., Chapman-Osterkatz, S., Niedzwiecki, S. and Shair-Rosenfield, S. (2016) Measuring regional Authority: A Postfunctional Theory of Governance,Volume I (Oxford: Oxford University Press).

Hough, D. and Jeffery, C. (eds) (2006) Devolution and Electoral Politics (Manchester: Manches-ter University Press).

Jeffery, C. (2010) ‘Multi-Level Party Competition in Federal and Regional States’. In Enderlein,H., Wälti, S. and Zürn, M. (eds) Handbook on Multi-Level Governance (Cheltenham: EdwardElgar), pp. 136–52.

Jeffery, C. and Wincott, D. (2010) ‘Beyond Methodological Nationalism: The Challenge ofTerritorial Politics’. In Hay, C. (ed.) New Directions in Political Science (Basingstoke:Palgrave Macmillan/Political Studies Association), pp. 167–88.



Jolly, S. (2007) ‘The Europhile Fringe? Regionalist Party Support for European Integration’.European Union Politics, Vol. 8, No. 1, pp. 109–30.

Keating, M. (1998) The New Regionalism in Western Europe (Cheltenham: Edward Elgar).Keating, M. (2009) ‘Putting European Political Science Back Together’. European Journal of

Political Science, Vol. 1, No. 2, pp. 297–316.Laitin, D. (2000) ‘What is a Language Community?’ American Journal of Political Science, Vol.

44, No. 1, pp. 142–55.Loughlin, J., Hendriks, F. and Lidström, A. (eds) (2011) The Oxford Handbook of Local and

Regional Democracy in Europe (Oxford: Oxford University Press).Marsh, M. and Mikhaylov, S. (2010) ‘European Parliament Elections and EU Governance’. Living

Reviews in European Governance, Vol. 5, No. 4.Massetti, E. (2009) ‘Explaining Regionalist Party Positioning in a Multi-Dimensional

Ideological Space: A Framework for Analysis’. Regional and Federal Studies, Vol. 19,No. 4/5, pp. 501–31.

Massetti, E. and Schakel, A.H. (2013) ‘IdeologyMatters:WhyDecentralisation Has aDifferentiatedEffect on Regionalist Parties’ Fortunes in Western Democracies’. European Journal of PoliticalResearch, Vol. 52, No. 6, pp. 797–821.

Massetti, E. and Schakel, A.H. (2017) ‘Decentralization Reforms and Regionalist Parties Strength:Accommodation, Empowerment or Both?’ Political Studies, Vol. 65, No. 2, pp. 432–51.

Matilla, M. (2003) ‘Why Bother? Determinants of Turnout in the European Elections’. ElectoralStudies, Vol. 22, No. 3, pp. 449–68.

McEwen, N. and Moreno, L. (eds) (2005) The Territorial Politics of Welfare (London: Routledge).Oppenhuis, E., van der Eijk, C. and Franklin, M. (1996) ‘The Party Context: Outcomes’. In

van der Eijk, C. and Franklin, M. (eds) Choosing Europe? The European Electorate andNational Politics in the Face of Union (Ann Arbor, MI: University of Michigan Press),pp. 287–306.

Pallarés, F. and Keating, M. (2003) ‘Multi-Level Electoral Competition. Regional Electionsand Party Systems in Spain’. European Urban and Regional Studies, Vol. 10, No. 3,pp. 239–55.

Reif, K. (1985) ‘Ten Second-Order Elections’. In Reif, K. (ed.) Ten Second-Order Elections(Aldershot: Gower), pp. 1–36.

Reif, K. and Schmitt, H. (1980) ‘Nine Second-Order National Elections –AConceptual Frameworkfor the Analysis of European Elections Results’. European Journal of Political Research,Vol. 8, No. 1, pp. 3–44.

Schakel, A.H. (2015) ‘How to Analyze Second-Order Election Effects? A Refined Second-OrderElection Model’. Comparative European Politics, Vol. 13, No. 6, pp. 636–55.

Schakel, A.H. (ed.) (2017) Regional and National Elections in Eastern Europe. Territoriality ofthe Vote in Ten Countries (Houndmills: Palgrave Macmillan).

Schakel, A.H. and Dandoy, R. (2013) ‘Intreoduction: Territoriality of the Vote. A Framework forAnalysis’’. In Dandoy, R. and Schakel, A.H. (eds) Regional and National Elections in WesternEurope. Territoriality of the Vote in the Thirteen Countries (Houndmills: Palgrave Macmillan),pp. 1–26.

Schakel, A.H. and Jeffery, C. (2013) ‘Are Regional Elections Really Second-Order?’ RegionalStudies, Vol. 47, No. 3, pp. 323–41.

Swenden, W. and Maddens, B. (2009) ‘Introduction. Territorial Party Politics in Western Europe:A Framework for Analysis’. In Swenden, W. and Maddens, B. (eds) Territorial Party Politicsin Western Europe (Houndmills: Palgrave Macmillan), pp. 1–30.

Szöcsik, E. (2013) ‘Minority Nationalist Parties and European Integration: Europeanists ParExcellence?’ Paper presented 13th Biennial Conference of the European Union StudiesAssociation, Baltimore, pp. 9–11 May.

Arjan H. Schakel704


Tronconi, F. and Roux, C. (2009) ‘The Political Systems of Italian Regions Between State-WideLogics and Increasing Differentiation’. Modern Italy, Vol. 14, No. 2, pp. 151–66.

Van der Eijk, C., Franklin, M. and Marsh, M. (1996) ‘What Voters Teach Us About Europe-WideElections: What Europe-Wide Elections Teach Us About Voters’. Electoral Studies, Vol. 15,No. 2, pp. 149–66.

Weber, T. (2007) ‘Campaign Effects and Second-Order Cycles. A Top-Down Approach ToEuropean Parliament Elections’. European Union Politics, Vol. 8, No. 4, pp. 509–36.



rethinking european elections: the importance of regional … · 2018. 5. 4. · 6 denmark...

Documents